1 //===-- OpenMP.cpp -- Open MP directive lowering --------------------------===// 2 // 3 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. 4 // See https://llvm.org/LICENSE.txt for license information. 5 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception 6 // 7 //===----------------------------------------------------------------------===// 8 // 9 // Coding style: https://mlir.llvm.org/getting_started/DeveloperGuide/ 10 // 11 //===----------------------------------------------------------------------===// 12 13 #include "flang/Lower/OpenMP.h" 14 #include "flang/Common/idioms.h" 15 #include "flang/Lower/Bridge.h" 16 #include "flang/Lower/ConvertExpr.h" 17 #include "flang/Lower/PFTBuilder.h" 18 #include "flang/Lower/StatementContext.h" 19 #include "flang/Lower/Todo.h" 20 #include "flang/Optimizer/Builder/BoxValue.h" 21 #include "flang/Optimizer/Builder/FIRBuilder.h" 22 #include "flang/Parser/parse-tree.h" 23 #include "flang/Semantics/tools.h" 24 #include "mlir/Dialect/OpenMP/OpenMPDialect.h" 25 #include "llvm/Frontend/OpenMP/OMPConstants.h" 26 27 using namespace mlir; 28 29 int64_t Fortran::lower::getCollapseValue( 30 const Fortran::parser::OmpClauseList &clauseList) { 31 for (const auto &clause : clauseList.v) { 32 if (const auto &collapseClause = 33 std::get_if<Fortran::parser::OmpClause::Collapse>(&clause.u)) { 34 const auto *expr = Fortran::semantics::GetExpr(collapseClause->v); 35 return Fortran::evaluate::ToInt64(*expr).value(); 36 } 37 } 38 return 1; 39 } 40 41 static const Fortran::parser::Name * 42 getDesignatorNameIfDataRef(const Fortran::parser::Designator &designator) { 43 const auto *dataRef = std::get_if<Fortran::parser::DataRef>(&designator.u); 44 return dataRef ? std::get_if<Fortran::parser::Name>(&dataRef->u) : nullptr; 45 } 46 47 static Fortran::semantics::Symbol * 48 getOmpObjectSymbol(const Fortran::parser::OmpObject &ompObject) { 49 Fortran::semantics::Symbol *sym = nullptr; 50 std::visit(Fortran::common::visitors{ 51 [&](const Fortran::parser::Designator &designator) { 52 if (const Fortran::parser::Name *name = 53 getDesignatorNameIfDataRef(designator)) { 54 sym = name->symbol; 55 } 56 }, 57 [&](const Fortran::parser::Name &name) { sym = name.symbol; }}, 58 ompObject.u); 59 return sym; 60 } 61 62 template <typename T> 63 static void createPrivateVarSyms(Fortran::lower::AbstractConverter &converter, 64 const T *clause) { 65 const Fortran::parser::OmpObjectList &ompObjectList = clause->v; 66 for (const Fortran::parser::OmpObject &ompObject : ompObjectList.v) { 67 Fortran::semantics::Symbol *sym = getOmpObjectSymbol(ompObject); 68 // Privatization for symbols which are pre-determined (like loop index 69 // variables) happen separately, for everything else privatize here 70 if constexpr (std::is_same_v<T, Fortran::parser::OmpClause::Firstprivate>) { 71 converter.copyHostAssociateVar(*sym); 72 } else { 73 bool success = converter.createHostAssociateVarClone(*sym); 74 (void)success; 75 assert(success && "Privatization failed due to existing binding"); 76 } 77 } 78 } 79 80 static void privatizeVars(Fortran::lower::AbstractConverter &converter, 81 const Fortran::parser::OmpClauseList &opClauseList) { 82 fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder(); 83 auto insPt = firOpBuilder.saveInsertionPoint(); 84 firOpBuilder.setInsertionPointToStart(firOpBuilder.getAllocaBlock()); 85 for (const Fortran::parser::OmpClause &clause : opClauseList.v) { 86 if (const auto &privateClause = 87 std::get_if<Fortran::parser::OmpClause::Private>(&clause.u)) { 88 createPrivateVarSyms(converter, privateClause); 89 } else if (const auto &firstPrivateClause = 90 std::get_if<Fortran::parser::OmpClause::Firstprivate>( 91 &clause.u)) { 92 createPrivateVarSyms(converter, firstPrivateClause); 93 } 94 } 95 firOpBuilder.restoreInsertionPoint(insPt); 96 } 97 98 /// The COMMON block is a global structure. \p commonValue is the base address 99 /// of the the COMMON block. As the offset from the symbol \p sym, generate the 100 /// COMMON block member value (commonValue + offset) for the symbol. 101 /// FIXME: Share the code with `instantiateCommon` in ConvertVariable.cpp. 102 static mlir::Value 103 genCommonBlockMember(Fortran::lower::AbstractConverter &converter, 104 const Fortran::semantics::Symbol &sym, 105 mlir::Value commonValue) { 106 auto &firOpBuilder = converter.getFirOpBuilder(); 107 mlir::Location currentLocation = converter.getCurrentLocation(); 108 mlir::IntegerType i8Ty = firOpBuilder.getIntegerType(8); 109 mlir::Type i8Ptr = firOpBuilder.getRefType(i8Ty); 110 mlir::Type seqTy = firOpBuilder.getRefType(firOpBuilder.getVarLenSeqTy(i8Ty)); 111 mlir::Value base = 112 firOpBuilder.createConvert(currentLocation, seqTy, commonValue); 113 std::size_t byteOffset = sym.GetUltimate().offset(); 114 mlir::Value offs = firOpBuilder.createIntegerConstant( 115 currentLocation, firOpBuilder.getIndexType(), byteOffset); 116 mlir::Value varAddr = firOpBuilder.create<fir::CoordinateOp>( 117 currentLocation, i8Ptr, base, mlir::ValueRange{offs}); 118 mlir::Type symType = converter.genType(sym); 119 return firOpBuilder.createConvert(currentLocation, 120 firOpBuilder.getRefType(symType), varAddr); 121 } 122 123 // Get the extended value for \p val by extracting additional variable 124 // information from \p base. 125 static fir::ExtendedValue getExtendedValue(fir::ExtendedValue base, 126 mlir::Value val) { 127 return base.match( 128 [&](const fir::MutableBoxValue &box) -> fir::ExtendedValue { 129 return fir::MutableBoxValue(val, box.nonDeferredLenParams(), {}); 130 }, 131 [&](const auto &) -> fir::ExtendedValue { 132 return fir::substBase(base, val); 133 }); 134 } 135 136 static void threadPrivatizeVars(Fortran::lower::AbstractConverter &converter, 137 Fortran::lower::pft::Evaluation &eval) { 138 auto &firOpBuilder = converter.getFirOpBuilder(); 139 mlir::Location currentLocation = converter.getCurrentLocation(); 140 auto insPt = firOpBuilder.saveInsertionPoint(); 141 firOpBuilder.setInsertionPointToStart(firOpBuilder.getAllocaBlock()); 142 143 // Get the original ThreadprivateOp corresponding to the symbol and use the 144 // symbol value from that opeartion to create one ThreadprivateOp copy 145 // operation inside the parallel region. 146 auto genThreadprivateOp = [&](Fortran::lower::SymbolRef sym) -> mlir::Value { 147 mlir::Value symOriThreadprivateValue = converter.getSymbolAddress(sym); 148 mlir::Operation *op = symOriThreadprivateValue.getDefiningOp(); 149 assert(mlir::isa<mlir::omp::ThreadprivateOp>(op) && 150 "The threadprivate operation not created"); 151 mlir::Value symValue = 152 mlir::dyn_cast<mlir::omp::ThreadprivateOp>(op).sym_addr(); 153 return firOpBuilder.create<mlir::omp::ThreadprivateOp>( 154 currentLocation, symValue.getType(), symValue); 155 }; 156 157 llvm::SetVector<const Fortran::semantics::Symbol *> threadprivateSyms; 158 converter.collectSymbolSet( 159 eval, threadprivateSyms, 160 Fortran::semantics::Symbol::Flag::OmpThreadprivate); 161 162 // For a COMMON block, the ThreadprivateOp is generated for itself instead of 163 // its members, so only bind the value of the new copied ThreadprivateOp 164 // inside the parallel region to the common block symbol only once for 165 // multiple members in one COMMON block. 166 llvm::SetVector<const Fortran::semantics::Symbol *> commonSyms; 167 for (std::size_t i = 0; i < threadprivateSyms.size(); i++) { 168 auto sym = threadprivateSyms[i]; 169 mlir::Value symThreadprivateValue; 170 if (const Fortran::semantics::Symbol *common = 171 Fortran::semantics::FindCommonBlockContaining(sym->GetUltimate())) { 172 mlir::Value commonThreadprivateValue; 173 if (commonSyms.contains(common)) { 174 commonThreadprivateValue = converter.getSymbolAddress(*common); 175 } else { 176 commonThreadprivateValue = genThreadprivateOp(*common); 177 converter.bindSymbol(*common, commonThreadprivateValue); 178 commonSyms.insert(common); 179 } 180 symThreadprivateValue = 181 genCommonBlockMember(converter, *sym, commonThreadprivateValue); 182 } else { 183 symThreadprivateValue = genThreadprivateOp(*sym); 184 } 185 186 fir::ExtendedValue sexv = converter.getSymbolExtendedValue(*sym); 187 fir::ExtendedValue symThreadprivateExv = 188 getExtendedValue(sexv, symThreadprivateValue); 189 converter.bindSymbol(*sym, symThreadprivateExv); 190 } 191 192 firOpBuilder.restoreInsertionPoint(insPt); 193 } 194 195 static void genObjectList(const Fortran::parser::OmpObjectList &objectList, 196 Fortran::lower::AbstractConverter &converter, 197 llvm::SmallVectorImpl<Value> &operands) { 198 auto addOperands = [&](Fortran::lower::SymbolRef sym) { 199 const mlir::Value variable = converter.getSymbolAddress(sym); 200 if (variable) { 201 operands.push_back(variable); 202 } else { 203 if (const auto *details = 204 sym->detailsIf<Fortran::semantics::HostAssocDetails>()) { 205 operands.push_back(converter.getSymbolAddress(details->symbol())); 206 converter.copySymbolBinding(details->symbol(), sym); 207 } 208 } 209 }; 210 for (const Fortran::parser::OmpObject &ompObject : objectList.v) { 211 Fortran::semantics::Symbol *sym = getOmpObjectSymbol(ompObject); 212 addOperands(*sym); 213 } 214 } 215 216 static mlir::Type getLoopVarType(Fortran::lower::AbstractConverter &converter, 217 std::size_t loopVarTypeSize) { 218 // OpenMP runtime requires 32-bit or 64-bit loop variables. 219 loopVarTypeSize = loopVarTypeSize * 8; 220 if (loopVarTypeSize < 32) { 221 loopVarTypeSize = 32; 222 } else if (loopVarTypeSize > 64) { 223 loopVarTypeSize = 64; 224 mlir::emitWarning(converter.getCurrentLocation(), 225 "OpenMP loop iteration variable cannot have more than 64 " 226 "bits size and will be narrowed into 64 bits."); 227 } 228 assert((loopVarTypeSize == 32 || loopVarTypeSize == 64) && 229 "OpenMP loop iteration variable size must be transformed into 32-bit " 230 "or 64-bit"); 231 return converter.getFirOpBuilder().getIntegerType(loopVarTypeSize); 232 } 233 234 /// Create empty blocks for the current region. 235 /// These blocks replace blocks parented to an enclosing region. 236 void createEmptyRegionBlocks( 237 fir::FirOpBuilder &firOpBuilder, 238 std::list<Fortran::lower::pft::Evaluation> &evaluationList) { 239 auto *region = &firOpBuilder.getRegion(); 240 for (auto &eval : evaluationList) { 241 if (eval.block) { 242 if (eval.block->empty()) { 243 eval.block->erase(); 244 eval.block = firOpBuilder.createBlock(region); 245 } else { 246 [[maybe_unused]] auto &terminatorOp = eval.block->back(); 247 assert((mlir::isa<mlir::omp::TerminatorOp>(terminatorOp) || 248 mlir::isa<mlir::omp::YieldOp>(terminatorOp)) && 249 "expected terminator op"); 250 } 251 } 252 if (!eval.isDirective() && eval.hasNestedEvaluations()) 253 createEmptyRegionBlocks(firOpBuilder, eval.getNestedEvaluations()); 254 } 255 } 256 257 /// Create the body (block) for an OpenMP Operation. 258 /// 259 /// \param [in] op - the operation the body belongs to. 260 /// \param [inout] converter - converter to use for the clauses. 261 /// \param [in] loc - location in source code. 262 /// \param [in] eval - current PFT node/evaluation. 263 /// \oaran [in] clauses - list of clauses to process. 264 /// \param [in] args - block arguments (induction variable[s]) for the 265 //// region. 266 /// \param [in] outerCombined - is this an outer operation - prevents 267 /// privatization. 268 template <typename Op> 269 static void 270 createBodyOfOp(Op &op, Fortran::lower::AbstractConverter &converter, 271 mlir::Location &loc, Fortran::lower::pft::Evaluation &eval, 272 const Fortran::parser::OmpClauseList *clauses = nullptr, 273 const SmallVector<const Fortran::semantics::Symbol *> &args = {}, 274 bool outerCombined = false) { 275 fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder(); 276 // If an argument for the region is provided then create the block with that 277 // argument. Also update the symbol's address with the mlir argument value. 278 // e.g. For loops the argument is the induction variable. And all further 279 // uses of the induction variable should use this mlir value. 280 mlir::Operation *storeOp = nullptr; 281 if (args.size()) { 282 std::size_t loopVarTypeSize = 0; 283 for (const Fortran::semantics::Symbol *arg : args) 284 loopVarTypeSize = std::max(loopVarTypeSize, arg->GetUltimate().size()); 285 mlir::Type loopVarType = getLoopVarType(converter, loopVarTypeSize); 286 SmallVector<Type> tiv; 287 SmallVector<Location> locs; 288 for (int i = 0; i < (int)args.size(); i++) { 289 tiv.push_back(loopVarType); 290 locs.push_back(loc); 291 } 292 firOpBuilder.createBlock(&op.getRegion(), {}, tiv, locs); 293 int argIndex = 0; 294 // The argument is not currently in memory, so make a temporary for the 295 // argument, and store it there, then bind that location to the argument. 296 for (const Fortran::semantics::Symbol *arg : args) { 297 mlir::Value val = 298 fir::getBase(op.getRegion().front().getArgument(argIndex)); 299 mlir::Value temp = firOpBuilder.createTemporary( 300 loc, loopVarType, 301 llvm::ArrayRef<mlir::NamedAttribute>{ 302 Fortran::lower::getAdaptToByRefAttr(firOpBuilder)}); 303 storeOp = firOpBuilder.create<fir::StoreOp>(loc, val, temp); 304 converter.bindSymbol(*arg, temp); 305 argIndex++; 306 } 307 } else { 308 firOpBuilder.createBlock(&op.getRegion()); 309 } 310 // Set the insert for the terminator operation to go at the end of the 311 // block - this is either empty or the block with the stores above, 312 // the end of the block works for both. 313 mlir::Block &block = op.getRegion().back(); 314 firOpBuilder.setInsertionPointToEnd(&block); 315 316 // If it is an unstructured region and is not the outer region of a combined 317 // construct, create empty blocks for all evaluations. 318 if (eval.lowerAsUnstructured() && !outerCombined) 319 createEmptyRegionBlocks(firOpBuilder, eval.getNestedEvaluations()); 320 321 // Insert the terminator. 322 if constexpr (std::is_same_v<Op, omp::WsLoopOp>) { 323 mlir::ValueRange results; 324 firOpBuilder.create<mlir::omp::YieldOp>(loc, results); 325 } else { 326 firOpBuilder.create<mlir::omp::TerminatorOp>(loc); 327 } 328 329 // Reset the insert point to before the terminator. 330 if (storeOp) 331 firOpBuilder.setInsertionPointAfter(storeOp); 332 else 333 firOpBuilder.setInsertionPointToStart(&block); 334 335 // Handle privatization. Do not privatize if this is the outer operation. 336 if (clauses && !outerCombined) 337 privatizeVars(converter, *clauses); 338 339 if (std::is_same_v<Op, omp::ParallelOp>) 340 threadPrivatizeVars(converter, eval); 341 } 342 343 static void genOMP(Fortran::lower::AbstractConverter &converter, 344 Fortran::lower::pft::Evaluation &eval, 345 const Fortran::parser::OpenMPSimpleStandaloneConstruct 346 &simpleStandaloneConstruct) { 347 const auto &directive = 348 std::get<Fortran::parser::OmpSimpleStandaloneDirective>( 349 simpleStandaloneConstruct.t); 350 switch (directive.v) { 351 default: 352 break; 353 case llvm::omp::Directive::OMPD_barrier: 354 converter.getFirOpBuilder().create<mlir::omp::BarrierOp>( 355 converter.getCurrentLocation()); 356 break; 357 case llvm::omp::Directive::OMPD_taskwait: 358 converter.getFirOpBuilder().create<mlir::omp::TaskwaitOp>( 359 converter.getCurrentLocation()); 360 break; 361 case llvm::omp::Directive::OMPD_taskyield: 362 converter.getFirOpBuilder().create<mlir::omp::TaskyieldOp>( 363 converter.getCurrentLocation()); 364 break; 365 case llvm::omp::Directive::OMPD_target_enter_data: 366 TODO(converter.getCurrentLocation(), "OMPD_target_enter_data"); 367 case llvm::omp::Directive::OMPD_target_exit_data: 368 TODO(converter.getCurrentLocation(), "OMPD_target_exit_data"); 369 case llvm::omp::Directive::OMPD_target_update: 370 TODO(converter.getCurrentLocation(), "OMPD_target_update"); 371 case llvm::omp::Directive::OMPD_ordered: 372 TODO(converter.getCurrentLocation(), "OMPD_ordered"); 373 } 374 } 375 376 static void 377 genAllocateClause(Fortran::lower::AbstractConverter &converter, 378 const Fortran::parser::OmpAllocateClause &ompAllocateClause, 379 SmallVector<Value> &allocatorOperands, 380 SmallVector<Value> &allocateOperands) { 381 auto &firOpBuilder = converter.getFirOpBuilder(); 382 auto currentLocation = converter.getCurrentLocation(); 383 Fortran::lower::StatementContext stmtCtx; 384 385 mlir::Value allocatorOperand; 386 const Fortran::parser::OmpObjectList &ompObjectList = 387 std::get<Fortran::parser::OmpObjectList>(ompAllocateClause.t); 388 const auto &allocatorValue = 389 std::get<std::optional<Fortran::parser::OmpAllocateClause::Allocator>>( 390 ompAllocateClause.t); 391 // Check if allocate clause has allocator specified. If so, add it 392 // to list of allocators, otherwise, add default allocator to 393 // list of allocators. 394 if (allocatorValue) { 395 allocatorOperand = fir::getBase(converter.genExprValue( 396 *Fortran::semantics::GetExpr(allocatorValue->v), stmtCtx)); 397 allocatorOperands.insert(allocatorOperands.end(), ompObjectList.v.size(), 398 allocatorOperand); 399 } else { 400 allocatorOperand = firOpBuilder.createIntegerConstant( 401 currentLocation, firOpBuilder.getI32Type(), 1); 402 allocatorOperands.insert(allocatorOperands.end(), ompObjectList.v.size(), 403 allocatorOperand); 404 } 405 genObjectList(ompObjectList, converter, allocateOperands); 406 } 407 408 static void 409 genOMP(Fortran::lower::AbstractConverter &converter, 410 Fortran::lower::pft::Evaluation &eval, 411 const Fortran::parser::OpenMPStandaloneConstruct &standaloneConstruct) { 412 std::visit( 413 Fortran::common::visitors{ 414 [&](const Fortran::parser::OpenMPSimpleStandaloneConstruct 415 &simpleStandaloneConstruct) { 416 genOMP(converter, eval, simpleStandaloneConstruct); 417 }, 418 [&](const Fortran::parser::OpenMPFlushConstruct &flushConstruct) { 419 SmallVector<Value, 4> operandRange; 420 if (const auto &ompObjectList = 421 std::get<std::optional<Fortran::parser::OmpObjectList>>( 422 flushConstruct.t)) 423 genObjectList(*ompObjectList, converter, operandRange); 424 const auto &memOrderClause = std::get<std::optional< 425 std::list<Fortran::parser::OmpMemoryOrderClause>>>( 426 flushConstruct.t); 427 if (memOrderClause.has_value() && memOrderClause->size() > 0) 428 TODO(converter.getCurrentLocation(), 429 "Handle OmpMemoryOrderClause"); 430 converter.getFirOpBuilder().create<mlir::omp::FlushOp>( 431 converter.getCurrentLocation(), operandRange); 432 }, 433 [&](const Fortran::parser::OpenMPCancelConstruct &cancelConstruct) { 434 TODO(converter.getCurrentLocation(), "OpenMPCancelConstruct"); 435 }, 436 [&](const Fortran::parser::OpenMPCancellationPointConstruct 437 &cancellationPointConstruct) { 438 TODO(converter.getCurrentLocation(), "OpenMPCancelConstruct"); 439 }, 440 }, 441 standaloneConstruct.u); 442 } 443 444 static omp::ClauseProcBindKindAttr genProcBindKindAttr( 445 fir::FirOpBuilder &firOpBuilder, 446 const Fortran::parser::OmpClause::ProcBind *procBindClause) { 447 omp::ClauseProcBindKind pbKind; 448 switch (procBindClause->v.v) { 449 case Fortran::parser::OmpProcBindClause::Type::Master: 450 pbKind = omp::ClauseProcBindKind::Master; 451 break; 452 case Fortran::parser::OmpProcBindClause::Type::Close: 453 pbKind = omp::ClauseProcBindKind::Close; 454 break; 455 case Fortran::parser::OmpProcBindClause::Type::Spread: 456 pbKind = omp::ClauseProcBindKind::Spread; 457 break; 458 case Fortran::parser::OmpProcBindClause::Type::Primary: 459 pbKind = omp::ClauseProcBindKind::Primary; 460 break; 461 } 462 return omp::ClauseProcBindKindAttr::get(firOpBuilder.getContext(), pbKind); 463 } 464 465 /* When parallel is used in a combined construct, then use this function to 466 * create the parallel operation. It handles the parallel specific clauses 467 * and leaves the rest for handling at the inner operations. 468 * TODO: Refactor clause handling 469 */ 470 template <typename Directive> 471 static void 472 createCombinedParallelOp(Fortran::lower::AbstractConverter &converter, 473 Fortran::lower::pft::Evaluation &eval, 474 const Directive &directive) { 475 fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder(); 476 mlir::Location currentLocation = converter.getCurrentLocation(); 477 Fortran::lower::StatementContext stmtCtx; 478 llvm::ArrayRef<mlir::Type> argTy; 479 mlir::Value ifClauseOperand, numThreadsClauseOperand; 480 SmallVector<Value> allocatorOperands, allocateOperands; 481 mlir::omp::ClauseProcBindKindAttr procBindKindAttr; 482 const auto &opClauseList = 483 std::get<Fortran::parser::OmpClauseList>(directive.t); 484 // TODO: Handle the following clauses 485 // 1. default 486 // 2. copyin 487 // Note: rest of the clauses are handled when the inner operation is created 488 for (const Fortran::parser::OmpClause &clause : opClauseList.v) { 489 if (const auto &ifClause = 490 std::get_if<Fortran::parser::OmpClause::If>(&clause.u)) { 491 auto &expr = std::get<Fortran::parser::ScalarLogicalExpr>(ifClause->v.t); 492 mlir::Value ifVal = fir::getBase( 493 converter.genExprValue(*Fortran::semantics::GetExpr(expr), stmtCtx)); 494 ifClauseOperand = firOpBuilder.createConvert( 495 currentLocation, firOpBuilder.getI1Type(), ifVal); 496 } else if (const auto &numThreadsClause = 497 std::get_if<Fortran::parser::OmpClause::NumThreads>( 498 &clause.u)) { 499 numThreadsClauseOperand = fir::getBase(converter.genExprValue( 500 *Fortran::semantics::GetExpr(numThreadsClause->v), stmtCtx)); 501 } else if (const auto &procBindClause = 502 std::get_if<Fortran::parser::OmpClause::ProcBind>( 503 &clause.u)) { 504 procBindKindAttr = genProcBindKindAttr(firOpBuilder, procBindClause); 505 } 506 } 507 // Create and insert the operation. 508 auto parallelOp = firOpBuilder.create<mlir::omp::ParallelOp>( 509 currentLocation, argTy, ifClauseOperand, numThreadsClauseOperand, 510 allocateOperands, allocatorOperands, /*reduction_vars=*/ValueRange(), 511 /*reductions=*/nullptr, procBindKindAttr); 512 513 createBodyOfOp<omp::ParallelOp>(parallelOp, converter, currentLocation, eval, 514 &opClauseList, /*iv=*/{}, 515 /*isCombined=*/true); 516 } 517 518 static void 519 genOMP(Fortran::lower::AbstractConverter &converter, 520 Fortran::lower::pft::Evaluation &eval, 521 const Fortran::parser::OpenMPBlockConstruct &blockConstruct) { 522 const auto &beginBlockDirective = 523 std::get<Fortran::parser::OmpBeginBlockDirective>(blockConstruct.t); 524 const auto &blockDirective = 525 std::get<Fortran::parser::OmpBlockDirective>(beginBlockDirective.t); 526 const auto &endBlockDirective = 527 std::get<Fortran::parser::OmpEndBlockDirective>(blockConstruct.t); 528 fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder(); 529 mlir::Location currentLocation = converter.getCurrentLocation(); 530 531 Fortran::lower::StatementContext stmtCtx; 532 llvm::ArrayRef<mlir::Type> argTy; 533 mlir::Value ifClauseOperand, numThreadsClauseOperand, finalClauseOperand, 534 priorityClauseOperand; 535 mlir::omp::ClauseProcBindKindAttr procBindKindAttr; 536 SmallVector<Value> allocateOperands, allocatorOperands; 537 mlir::UnitAttr nowaitAttr, untiedAttr, mergeableAttr; 538 539 const auto &opClauseList = 540 std::get<Fortran::parser::OmpClauseList>(beginBlockDirective.t); 541 for (const auto &clause : opClauseList.v) { 542 if (const auto &ifClause = 543 std::get_if<Fortran::parser::OmpClause::If>(&clause.u)) { 544 auto &expr = std::get<Fortran::parser::ScalarLogicalExpr>(ifClause->v.t); 545 mlir::Value ifVal = fir::getBase( 546 converter.genExprValue(*Fortran::semantics::GetExpr(expr), stmtCtx)); 547 ifClauseOperand = firOpBuilder.createConvert( 548 currentLocation, firOpBuilder.getI1Type(), ifVal); 549 } else if (const auto &numThreadsClause = 550 std::get_if<Fortran::parser::OmpClause::NumThreads>( 551 &clause.u)) { 552 // OMPIRBuilder expects `NUM_THREAD` clause as a `Value`. 553 numThreadsClauseOperand = fir::getBase(converter.genExprValue( 554 *Fortran::semantics::GetExpr(numThreadsClause->v), stmtCtx)); 555 } else if (const auto &procBindClause = 556 std::get_if<Fortran::parser::OmpClause::ProcBind>( 557 &clause.u)) { 558 procBindKindAttr = genProcBindKindAttr(firOpBuilder, procBindClause); 559 } else if (const auto &allocateClause = 560 std::get_if<Fortran::parser::OmpClause::Allocate>( 561 &clause.u)) { 562 genAllocateClause(converter, allocateClause->v, allocatorOperands, 563 allocateOperands); 564 } else if (std::get_if<Fortran::parser::OmpClause::Private>(&clause.u) || 565 std::get_if<Fortran::parser::OmpClause::Firstprivate>( 566 &clause.u)) { 567 // Privatisation clauses are handled elsewhere. 568 continue; 569 } else if (std::get_if<Fortran::parser::OmpClause::Threads>(&clause.u)) { 570 // Nothing needs to be done for threads clause. 571 continue; 572 } else if (const auto &finalClause = 573 std::get_if<Fortran::parser::OmpClause::Final>(&clause.u)) { 574 mlir::Value finalVal = fir::getBase(converter.genExprValue( 575 *Fortran::semantics::GetExpr(finalClause->v), stmtCtx)); 576 finalClauseOperand = firOpBuilder.createConvert( 577 currentLocation, firOpBuilder.getI1Type(), finalVal); 578 } else if (std::get_if<Fortran::parser::OmpClause::Untied>(&clause.u)) { 579 untiedAttr = firOpBuilder.getUnitAttr(); 580 } else if (std::get_if<Fortran::parser::OmpClause::Mergeable>(&clause.u)) { 581 mergeableAttr = firOpBuilder.getUnitAttr(); 582 } else if (const auto &priorityClause = 583 std::get_if<Fortran::parser::OmpClause::Priority>( 584 &clause.u)) { 585 priorityClauseOperand = fir::getBase(converter.genExprValue( 586 *Fortran::semantics::GetExpr(priorityClause->v), stmtCtx)); 587 } else { 588 TODO(currentLocation, "OpenMP Block construct clauses"); 589 } 590 } 591 592 for (const auto &clause : 593 std::get<Fortran::parser::OmpClauseList>(endBlockDirective.t).v) { 594 if (std::get_if<Fortran::parser::OmpClause::Nowait>(&clause.u)) 595 nowaitAttr = firOpBuilder.getUnitAttr(); 596 } 597 598 if (blockDirective.v == llvm::omp::OMPD_parallel) { 599 // Create and insert the operation. 600 auto parallelOp = firOpBuilder.create<mlir::omp::ParallelOp>( 601 currentLocation, argTy, ifClauseOperand, numThreadsClauseOperand, 602 allocateOperands, allocatorOperands, /*reduction_vars=*/ValueRange(), 603 /*reductions=*/nullptr, procBindKindAttr); 604 createBodyOfOp<omp::ParallelOp>(parallelOp, converter, currentLocation, 605 eval, &opClauseList); 606 } else if (blockDirective.v == llvm::omp::OMPD_master) { 607 auto masterOp = 608 firOpBuilder.create<mlir::omp::MasterOp>(currentLocation, argTy); 609 createBodyOfOp<omp::MasterOp>(masterOp, converter, currentLocation, eval); 610 } else if (blockDirective.v == llvm::omp::OMPD_single) { 611 auto singleOp = firOpBuilder.create<mlir::omp::SingleOp>( 612 currentLocation, allocateOperands, allocatorOperands, nowaitAttr); 613 createBodyOfOp<omp::SingleOp>(singleOp, converter, currentLocation, eval); 614 } else if (blockDirective.v == llvm::omp::OMPD_ordered) { 615 auto orderedOp = firOpBuilder.create<mlir::omp::OrderedRegionOp>( 616 currentLocation, /*simd=*/nullptr); 617 createBodyOfOp<omp::OrderedRegionOp>(orderedOp, converter, currentLocation, 618 eval); 619 } else if (blockDirective.v == llvm::omp::OMPD_task) { 620 auto taskOp = firOpBuilder.create<mlir::omp::TaskOp>( 621 currentLocation, ifClauseOperand, finalClauseOperand, untiedAttr, 622 mergeableAttr, /*in_reduction_vars=*/ValueRange(), 623 /*in_reductions=*/nullptr, priorityClauseOperand, allocateOperands, 624 allocatorOperands); 625 createBodyOfOp(taskOp, converter, currentLocation, eval, &opClauseList); 626 } else { 627 TODO(converter.getCurrentLocation(), "Unhandled block directive"); 628 } 629 } 630 631 static mlir::omp::ScheduleModifier 632 translateModifier(const Fortran::parser::OmpScheduleModifierType &m) { 633 switch (m.v) { 634 case Fortran::parser::OmpScheduleModifierType::ModType::Monotonic: 635 return mlir::omp::ScheduleModifier::monotonic; 636 case Fortran::parser::OmpScheduleModifierType::ModType::Nonmonotonic: 637 return mlir::omp::ScheduleModifier::nonmonotonic; 638 case Fortran::parser::OmpScheduleModifierType::ModType::Simd: 639 return mlir::omp::ScheduleModifier::simd; 640 } 641 return mlir::omp::ScheduleModifier::none; 642 } 643 644 static mlir::omp::ScheduleModifier 645 getScheduleModifier(const Fortran::parser::OmpScheduleClause &x) { 646 const auto &modifier = 647 std::get<std::optional<Fortran::parser::OmpScheduleModifier>>(x.t); 648 // The input may have the modifier any order, so we look for one that isn't 649 // SIMD. If modifier is not set at all, fall down to the bottom and return 650 // "none". 651 if (modifier) { 652 const auto &modType1 = 653 std::get<Fortran::parser::OmpScheduleModifier::Modifier1>(modifier->t); 654 if (modType1.v.v == 655 Fortran::parser::OmpScheduleModifierType::ModType::Simd) { 656 const auto &modType2 = std::get< 657 std::optional<Fortran::parser::OmpScheduleModifier::Modifier2>>( 658 modifier->t); 659 if (modType2 && 660 modType2->v.v != 661 Fortran::parser::OmpScheduleModifierType::ModType::Simd) 662 return translateModifier(modType2->v); 663 664 return mlir::omp::ScheduleModifier::none; 665 } 666 667 return translateModifier(modType1.v); 668 } 669 return mlir::omp::ScheduleModifier::none; 670 } 671 672 static mlir::omp::ScheduleModifier 673 getSIMDModifier(const Fortran::parser::OmpScheduleClause &x) { 674 const auto &modifier = 675 std::get<std::optional<Fortran::parser::OmpScheduleModifier>>(x.t); 676 // Either of the two possible modifiers in the input can be the SIMD modifier, 677 // so look in either one, and return simd if we find one. Not found = return 678 // "none". 679 if (modifier) { 680 const auto &modType1 = 681 std::get<Fortran::parser::OmpScheduleModifier::Modifier1>(modifier->t); 682 if (modType1.v.v == Fortran::parser::OmpScheduleModifierType::ModType::Simd) 683 return mlir::omp::ScheduleModifier::simd; 684 685 const auto &modType2 = std::get< 686 std::optional<Fortran::parser::OmpScheduleModifier::Modifier2>>( 687 modifier->t); 688 if (modType2 && modType2->v.v == 689 Fortran::parser::OmpScheduleModifierType::ModType::Simd) 690 return mlir::omp::ScheduleModifier::simd; 691 } 692 return mlir::omp::ScheduleModifier::none; 693 } 694 695 static void genOMP(Fortran::lower::AbstractConverter &converter, 696 Fortran::lower::pft::Evaluation &eval, 697 const Fortran::parser::OpenMPLoopConstruct &loopConstruct) { 698 699 fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder(); 700 mlir::Location currentLocation = converter.getCurrentLocation(); 701 llvm::SmallVector<mlir::Value> lowerBound, upperBound, step, linearVars, 702 linearStepVars, reductionVars; 703 mlir::Value scheduleChunkClauseOperand; 704 mlir::Attribute scheduleClauseOperand, collapseClauseOperand, 705 noWaitClauseOperand, orderedClauseOperand, orderClauseOperand; 706 const auto &wsLoopOpClauseList = std::get<Fortran::parser::OmpClauseList>( 707 std::get<Fortran::parser::OmpBeginLoopDirective>(loopConstruct.t).t); 708 709 const auto ompDirective = 710 std::get<Fortran::parser::OmpLoopDirective>( 711 std::get<Fortran::parser::OmpBeginLoopDirective>(loopConstruct.t).t) 712 .v; 713 if (llvm::omp::OMPD_parallel_do == ompDirective) { 714 createCombinedParallelOp<Fortran::parser::OmpBeginLoopDirective>( 715 converter, eval, 716 std::get<Fortran::parser::OmpBeginLoopDirective>(loopConstruct.t)); 717 } else if (llvm::omp::OMPD_do != ompDirective) { 718 TODO(converter.getCurrentLocation(), "Construct enclosing do loop"); 719 } 720 721 // Collect the loops to collapse. 722 auto *doConstructEval = &eval.getFirstNestedEvaluation(); 723 724 std::int64_t collapseValue = 725 Fortran::lower::getCollapseValue(wsLoopOpClauseList); 726 std::size_t loopVarTypeSize = 0; 727 SmallVector<const Fortran::semantics::Symbol *> iv; 728 do { 729 auto *doLoop = &doConstructEval->getFirstNestedEvaluation(); 730 auto *doStmt = doLoop->getIf<Fortran::parser::NonLabelDoStmt>(); 731 assert(doStmt && "Expected do loop to be in the nested evaluation"); 732 const auto &loopControl = 733 std::get<std::optional<Fortran::parser::LoopControl>>(doStmt->t); 734 const Fortran::parser::LoopControl::Bounds *bounds = 735 std::get_if<Fortran::parser::LoopControl::Bounds>(&loopControl->u); 736 assert(bounds && "Expected bounds for worksharing do loop"); 737 Fortran::lower::StatementContext stmtCtx; 738 lowerBound.push_back(fir::getBase(converter.genExprValue( 739 *Fortran::semantics::GetExpr(bounds->lower), stmtCtx))); 740 upperBound.push_back(fir::getBase(converter.genExprValue( 741 *Fortran::semantics::GetExpr(bounds->upper), stmtCtx))); 742 if (bounds->step) { 743 step.push_back(fir::getBase(converter.genExprValue( 744 *Fortran::semantics::GetExpr(bounds->step), stmtCtx))); 745 } else { // If `step` is not present, assume it as `1`. 746 step.push_back(firOpBuilder.createIntegerConstant( 747 currentLocation, firOpBuilder.getIntegerType(32), 1)); 748 } 749 iv.push_back(bounds->name.thing.symbol); 750 loopVarTypeSize = std::max(loopVarTypeSize, 751 bounds->name.thing.symbol->GetUltimate().size()); 752 753 collapseValue--; 754 doConstructEval = 755 &*std::next(doConstructEval->getNestedEvaluations().begin()); 756 } while (collapseValue > 0); 757 758 for (const auto &clause : wsLoopOpClauseList.v) { 759 if (const auto &scheduleClause = 760 std::get_if<Fortran::parser::OmpClause::Schedule>(&clause.u)) { 761 if (const auto &chunkExpr = 762 std::get<std::optional<Fortran::parser::ScalarIntExpr>>( 763 scheduleClause->v.t)) { 764 if (const auto *expr = Fortran::semantics::GetExpr(*chunkExpr)) { 765 Fortran::lower::StatementContext stmtCtx; 766 scheduleChunkClauseOperand = 767 fir::getBase(converter.genExprValue(*expr, stmtCtx)); 768 } 769 } 770 } 771 } 772 773 // The types of lower bound, upper bound, and step are converted into the 774 // type of the loop variable if necessary. 775 mlir::Type loopVarType = getLoopVarType(converter, loopVarTypeSize); 776 for (unsigned it = 0; it < (unsigned)lowerBound.size(); it++) { 777 lowerBound[it] = firOpBuilder.createConvert(currentLocation, loopVarType, 778 lowerBound[it]); 779 upperBound[it] = firOpBuilder.createConvert(currentLocation, loopVarType, 780 upperBound[it]); 781 step[it] = 782 firOpBuilder.createConvert(currentLocation, loopVarType, step[it]); 783 } 784 785 // FIXME: Add support for following clauses: 786 // 1. linear 787 // 2. order 788 auto wsLoopOp = firOpBuilder.create<mlir::omp::WsLoopOp>( 789 currentLocation, lowerBound, upperBound, step, linearVars, linearStepVars, 790 reductionVars, /*reductions=*/nullptr, 791 scheduleClauseOperand.dyn_cast_or_null<omp::ClauseScheduleKindAttr>(), 792 scheduleChunkClauseOperand, /*schedule_modifiers=*/nullptr, 793 /*simd_modifier=*/nullptr, 794 collapseClauseOperand.dyn_cast_or_null<IntegerAttr>(), 795 noWaitClauseOperand.dyn_cast_or_null<UnitAttr>(), 796 orderedClauseOperand.dyn_cast_or_null<IntegerAttr>(), 797 orderClauseOperand.dyn_cast_or_null<omp::ClauseOrderKindAttr>(), 798 /*inclusive=*/firOpBuilder.getUnitAttr()); 799 800 // Handle attribute based clauses. 801 for (const Fortran::parser::OmpClause &clause : wsLoopOpClauseList.v) { 802 if (const auto &orderedClause = 803 std::get_if<Fortran::parser::OmpClause::Ordered>(&clause.u)) { 804 if (orderedClause->v.has_value()) { 805 const auto *expr = Fortran::semantics::GetExpr(orderedClause->v); 806 const std::optional<std::int64_t> orderedClauseValue = 807 Fortran::evaluate::ToInt64(*expr); 808 wsLoopOp.ordered_valAttr( 809 firOpBuilder.getI64IntegerAttr(*orderedClauseValue)); 810 } else { 811 wsLoopOp.ordered_valAttr(firOpBuilder.getI64IntegerAttr(0)); 812 } 813 } else if (const auto &collapseClause = 814 std::get_if<Fortran::parser::OmpClause::Collapse>( 815 &clause.u)) { 816 const auto *expr = Fortran::semantics::GetExpr(collapseClause->v); 817 const std::optional<std::int64_t> collapseValue = 818 Fortran::evaluate::ToInt64(*expr); 819 wsLoopOp.collapse_valAttr(firOpBuilder.getI64IntegerAttr(*collapseValue)); 820 } else if (const auto &scheduleClause = 821 std::get_if<Fortran::parser::OmpClause::Schedule>( 822 &clause.u)) { 823 mlir::MLIRContext *context = firOpBuilder.getContext(); 824 const auto &scheduleType = scheduleClause->v; 825 const auto &scheduleKind = 826 std::get<Fortran::parser::OmpScheduleClause::ScheduleType>( 827 scheduleType.t); 828 switch (scheduleKind) { 829 case Fortran::parser::OmpScheduleClause::ScheduleType::Static: 830 wsLoopOp.schedule_valAttr(omp::ClauseScheduleKindAttr::get( 831 context, omp::ClauseScheduleKind::Static)); 832 break; 833 case Fortran::parser::OmpScheduleClause::ScheduleType::Dynamic: 834 wsLoopOp.schedule_valAttr(omp::ClauseScheduleKindAttr::get( 835 context, omp::ClauseScheduleKind::Dynamic)); 836 break; 837 case Fortran::parser::OmpScheduleClause::ScheduleType::Guided: 838 wsLoopOp.schedule_valAttr(omp::ClauseScheduleKindAttr::get( 839 context, omp::ClauseScheduleKind::Guided)); 840 break; 841 case Fortran::parser::OmpScheduleClause::ScheduleType::Auto: 842 wsLoopOp.schedule_valAttr(omp::ClauseScheduleKindAttr::get( 843 context, omp::ClauseScheduleKind::Auto)); 844 break; 845 case Fortran::parser::OmpScheduleClause::ScheduleType::Runtime: 846 wsLoopOp.schedule_valAttr(omp::ClauseScheduleKindAttr::get( 847 context, omp::ClauseScheduleKind::Runtime)); 848 break; 849 } 850 mlir::omp::ScheduleModifier scheduleModifier = 851 getScheduleModifier(scheduleClause->v); 852 if (scheduleModifier != mlir::omp::ScheduleModifier::none) 853 wsLoopOp.schedule_modifierAttr( 854 omp::ScheduleModifierAttr::get(context, scheduleModifier)); 855 if (getSIMDModifier(scheduleClause->v) != 856 mlir::omp::ScheduleModifier::none) 857 wsLoopOp.simd_modifierAttr(firOpBuilder.getUnitAttr()); 858 } 859 } 860 // In FORTRAN `nowait` clause occur at the end of `omp do` directive. 861 // i.e 862 // !$omp do 863 // <...> 864 // !$omp end do nowait 865 if (const auto &endClauseList = 866 std::get<std::optional<Fortran::parser::OmpEndLoopDirective>>( 867 loopConstruct.t)) { 868 const auto &clauseList = 869 std::get<Fortran::parser::OmpClauseList>((*endClauseList).t); 870 for (const Fortran::parser::OmpClause &clause : clauseList.v) 871 if (std::get_if<Fortran::parser::OmpClause::Nowait>(&clause.u)) 872 wsLoopOp.nowaitAttr(firOpBuilder.getUnitAttr()); 873 } 874 875 createBodyOfOp<omp::WsLoopOp>(wsLoopOp, converter, currentLocation, eval, 876 &wsLoopOpClauseList, iv); 877 } 878 879 static void 880 genOMP(Fortran::lower::AbstractConverter &converter, 881 Fortran::lower::pft::Evaluation &eval, 882 const Fortran::parser::OpenMPCriticalConstruct &criticalConstruct) { 883 fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder(); 884 mlir::Location currentLocation = converter.getCurrentLocation(); 885 std::string name; 886 const Fortran::parser::OmpCriticalDirective &cd = 887 std::get<Fortran::parser::OmpCriticalDirective>(criticalConstruct.t); 888 if (std::get<std::optional<Fortran::parser::Name>>(cd.t).has_value()) { 889 name = 890 std::get<std::optional<Fortran::parser::Name>>(cd.t).value().ToString(); 891 } 892 893 uint64_t hint = 0; 894 const auto &clauseList = std::get<Fortran::parser::OmpClauseList>(cd.t); 895 for (const Fortran::parser::OmpClause &clause : clauseList.v) 896 if (auto hintClause = 897 std::get_if<Fortran::parser::OmpClause::Hint>(&clause.u)) { 898 const auto *expr = Fortran::semantics::GetExpr(hintClause->v); 899 hint = *Fortran::evaluate::ToInt64(*expr); 900 break; 901 } 902 903 mlir::omp::CriticalOp criticalOp = [&]() { 904 if (name.empty()) { 905 return firOpBuilder.create<mlir::omp::CriticalOp>(currentLocation, 906 FlatSymbolRefAttr()); 907 } else { 908 mlir::ModuleOp module = firOpBuilder.getModule(); 909 mlir::OpBuilder modBuilder(module.getBodyRegion()); 910 auto global = module.lookupSymbol<mlir::omp::CriticalDeclareOp>(name); 911 if (!global) 912 global = modBuilder.create<mlir::omp::CriticalDeclareOp>( 913 currentLocation, name, hint); 914 return firOpBuilder.create<mlir::omp::CriticalOp>( 915 currentLocation, mlir::FlatSymbolRefAttr::get( 916 firOpBuilder.getContext(), global.sym_name())); 917 } 918 }(); 919 createBodyOfOp<omp::CriticalOp>(criticalOp, converter, currentLocation, eval); 920 } 921 922 static void 923 genOMP(Fortran::lower::AbstractConverter &converter, 924 Fortran::lower::pft::Evaluation &eval, 925 const Fortran::parser::OpenMPSectionConstruct §ionConstruct) { 926 927 auto &firOpBuilder = converter.getFirOpBuilder(); 928 auto currentLocation = converter.getCurrentLocation(); 929 mlir::omp::SectionOp sectionOp = 930 firOpBuilder.create<mlir::omp::SectionOp>(currentLocation); 931 createBodyOfOp<omp::SectionOp>(sectionOp, converter, currentLocation, eval); 932 } 933 934 // TODO: Add support for reduction 935 static void 936 genOMP(Fortran::lower::AbstractConverter &converter, 937 Fortran::lower::pft::Evaluation &eval, 938 const Fortran::parser::OpenMPSectionsConstruct §ionsConstruct) { 939 auto &firOpBuilder = converter.getFirOpBuilder(); 940 auto currentLocation = converter.getCurrentLocation(); 941 SmallVector<Value> reductionVars, allocateOperands, allocatorOperands; 942 mlir::UnitAttr noWaitClauseOperand; 943 const auto §ionsClauseList = std::get<Fortran::parser::OmpClauseList>( 944 std::get<Fortran::parser::OmpBeginSectionsDirective>(sectionsConstruct.t) 945 .t); 946 for (const Fortran::parser::OmpClause &clause : sectionsClauseList.v) { 947 948 // Reduction Clause 949 if (std::get_if<Fortran::parser::OmpClause::Reduction>(&clause.u)) { 950 TODO(currentLocation, "OMPC_Reduction"); 951 952 // Allocate clause 953 } else if (const auto &allocateClause = 954 std::get_if<Fortran::parser::OmpClause::Allocate>( 955 &clause.u)) { 956 genAllocateClause(converter, allocateClause->v, allocatorOperands, 957 allocateOperands); 958 } 959 } 960 const auto &endSectionsClauseList = 961 std::get<Fortran::parser::OmpEndSectionsDirective>(sectionsConstruct.t); 962 const auto &clauseList = 963 std::get<Fortran::parser::OmpClauseList>(endSectionsClauseList.t); 964 for (const auto &clause : clauseList.v) { 965 // Nowait clause 966 if (std::get_if<Fortran::parser::OmpClause::Nowait>(&clause.u)) { 967 noWaitClauseOperand = firOpBuilder.getUnitAttr(); 968 } 969 } 970 971 llvm::omp::Directive dir = 972 std::get<Fortran::parser::OmpSectionsDirective>( 973 std::get<Fortran::parser::OmpBeginSectionsDirective>( 974 sectionsConstruct.t) 975 .t) 976 .v; 977 978 // Parallel Sections Construct 979 if (dir == llvm::omp::Directive::OMPD_parallel_sections) { 980 createCombinedParallelOp<Fortran::parser::OmpBeginSectionsDirective>( 981 converter, eval, 982 std::get<Fortran::parser::OmpBeginSectionsDirective>( 983 sectionsConstruct.t)); 984 auto sectionsOp = firOpBuilder.create<mlir::omp::SectionsOp>( 985 currentLocation, /*reduction_vars*/ ValueRange(), 986 /*reductions=*/nullptr, allocateOperands, allocatorOperands, 987 /*nowait=*/nullptr); 988 createBodyOfOp(sectionsOp, converter, currentLocation, eval); 989 990 // Sections Construct 991 } else if (dir == llvm::omp::Directive::OMPD_sections) { 992 auto sectionsOp = firOpBuilder.create<mlir::omp::SectionsOp>( 993 currentLocation, reductionVars, /*reductions = */ nullptr, 994 allocateOperands, allocatorOperands, noWaitClauseOperand); 995 createBodyOfOp<omp::SectionsOp>(sectionsOp, converter, currentLocation, 996 eval); 997 } 998 } 999 1000 static void genOmpAtomicHintAndMemoryOrderClauses( 1001 Fortran::lower::AbstractConverter &converter, 1002 const Fortran::parser::OmpAtomicClauseList &clauseList, 1003 mlir::IntegerAttr &hint, 1004 mlir::omp::ClauseMemoryOrderKindAttr &memory_order) { 1005 auto &firOpBuilder = converter.getFirOpBuilder(); 1006 for (const auto &clause : clauseList.v) { 1007 if (auto ompClause = std::get_if<Fortran::parser::OmpClause>(&clause.u)) { 1008 if (auto hintClause = 1009 std::get_if<Fortran::parser::OmpClause::Hint>(&ompClause->u)) { 1010 const auto *expr = Fortran::semantics::GetExpr(hintClause->v); 1011 uint64_t hintExprValue = *Fortran::evaluate::ToInt64(*expr); 1012 hint = firOpBuilder.getI64IntegerAttr(hintExprValue); 1013 } 1014 } else if (auto ompMemoryOrderClause = 1015 std::get_if<Fortran::parser::OmpMemoryOrderClause>( 1016 &clause.u)) { 1017 if (std::get_if<Fortran::parser::OmpClause::Acquire>( 1018 &ompMemoryOrderClause->v.u)) { 1019 memory_order = mlir::omp::ClauseMemoryOrderKindAttr::get( 1020 firOpBuilder.getContext(), omp::ClauseMemoryOrderKind::Acquire); 1021 } else if (std::get_if<Fortran::parser::OmpClause::Relaxed>( 1022 &ompMemoryOrderClause->v.u)) { 1023 memory_order = mlir::omp::ClauseMemoryOrderKindAttr::get( 1024 firOpBuilder.getContext(), omp::ClauseMemoryOrderKind::Relaxed); 1025 } else if (std::get_if<Fortran::parser::OmpClause::SeqCst>( 1026 &ompMemoryOrderClause->v.u)) { 1027 memory_order = mlir::omp::ClauseMemoryOrderKindAttr::get( 1028 firOpBuilder.getContext(), omp::ClauseMemoryOrderKind::Seq_cst); 1029 } else if (std::get_if<Fortran::parser::OmpClause::Release>( 1030 &ompMemoryOrderClause->v.u)) { 1031 memory_order = mlir::omp::ClauseMemoryOrderKindAttr::get( 1032 firOpBuilder.getContext(), omp::ClauseMemoryOrderKind::Release); 1033 } 1034 } 1035 } 1036 } 1037 1038 static void 1039 genOmpAtomicWrite(Fortran::lower::AbstractConverter &converter, 1040 Fortran::lower::pft::Evaluation &eval, 1041 const Fortran::parser::OmpAtomicWrite &atomicWrite) { 1042 auto &firOpBuilder = converter.getFirOpBuilder(); 1043 auto currentLocation = converter.getCurrentLocation(); 1044 // Get the value and address of atomic write operands. 1045 const Fortran::parser::OmpAtomicClauseList &rightHandClauseList = 1046 std::get<2>(atomicWrite.t); 1047 const Fortran::parser::OmpAtomicClauseList &leftHandClauseList = 1048 std::get<0>(atomicWrite.t); 1049 const auto &assignmentStmtExpr = 1050 std::get<Fortran::parser::Expr>(std::get<3>(atomicWrite.t).statement.t); 1051 const auto &assignmentStmtVariable = std::get<Fortran::parser::Variable>( 1052 std::get<3>(atomicWrite.t).statement.t); 1053 Fortran::lower::StatementContext stmtCtx; 1054 mlir::Value value = fir::getBase(converter.genExprValue( 1055 *Fortran::semantics::GetExpr(assignmentStmtExpr), stmtCtx)); 1056 mlir::Value address = fir::getBase(converter.genExprAddr( 1057 *Fortran::semantics::GetExpr(assignmentStmtVariable), stmtCtx)); 1058 // If no hint clause is specified, the effect is as if 1059 // hint(omp_sync_hint_none) had been specified. 1060 mlir::IntegerAttr hint = nullptr; 1061 mlir::omp::ClauseMemoryOrderKindAttr memory_order = nullptr; 1062 genOmpAtomicHintAndMemoryOrderClauses(converter, leftHandClauseList, hint, 1063 memory_order); 1064 genOmpAtomicHintAndMemoryOrderClauses(converter, rightHandClauseList, hint, 1065 memory_order); 1066 firOpBuilder.create<mlir::omp::AtomicWriteOp>(currentLocation, address, value, 1067 hint, memory_order); 1068 } 1069 1070 static void genOmpAtomicRead(Fortran::lower::AbstractConverter &converter, 1071 Fortran::lower::pft::Evaluation &eval, 1072 const Fortran::parser::OmpAtomicRead &atomicRead) { 1073 auto &firOpBuilder = converter.getFirOpBuilder(); 1074 auto currentLocation = converter.getCurrentLocation(); 1075 // Get the address of atomic read operands. 1076 const Fortran::parser::OmpAtomicClauseList &rightHandClauseList = 1077 std::get<2>(atomicRead.t); 1078 const Fortran::parser::OmpAtomicClauseList &leftHandClauseList = 1079 std::get<0>(atomicRead.t); 1080 const auto &assignmentStmtExpr = 1081 std::get<Fortran::parser::Expr>(std::get<3>(atomicRead.t).statement.t); 1082 const auto &assignmentStmtVariable = std::get<Fortran::parser::Variable>( 1083 std::get<3>(atomicRead.t).statement.t); 1084 Fortran::lower::StatementContext stmtCtx; 1085 mlir::Value from_address = fir::getBase(converter.genExprAddr( 1086 *Fortran::semantics::GetExpr(assignmentStmtExpr), stmtCtx)); 1087 mlir::Value to_address = fir::getBase(converter.genExprAddr( 1088 *Fortran::semantics::GetExpr(assignmentStmtVariable), stmtCtx)); 1089 // If no hint clause is specified, the effect is as if 1090 // hint(omp_sync_hint_none) had been specified. 1091 mlir::IntegerAttr hint = nullptr; 1092 mlir::omp::ClauseMemoryOrderKindAttr memory_order = nullptr; 1093 genOmpAtomicHintAndMemoryOrderClauses(converter, leftHandClauseList, hint, 1094 memory_order); 1095 genOmpAtomicHintAndMemoryOrderClauses(converter, rightHandClauseList, hint, 1096 memory_order); 1097 firOpBuilder.create<mlir::omp::AtomicReadOp>(currentLocation, from_address, 1098 to_address, hint, memory_order); 1099 } 1100 1101 static void 1102 genOMP(Fortran::lower::AbstractConverter &converter, 1103 Fortran::lower::pft::Evaluation &eval, 1104 const Fortran::parser::OpenMPAtomicConstruct &atomicConstruct) { 1105 std::visit(Fortran::common::visitors{ 1106 [&](const Fortran::parser::OmpAtomicRead &atomicRead) { 1107 genOmpAtomicRead(converter, eval, atomicRead); 1108 }, 1109 [&](const Fortran::parser::OmpAtomicWrite &atomicWrite) { 1110 genOmpAtomicWrite(converter, eval, atomicWrite); 1111 }, 1112 [&](const auto &) { 1113 TODO(converter.getCurrentLocation(), 1114 "Atomic update & capture"); 1115 }, 1116 }, 1117 atomicConstruct.u); 1118 } 1119 1120 void Fortran::lower::genOpenMPConstruct( 1121 Fortran::lower::AbstractConverter &converter, 1122 Fortran::lower::pft::Evaluation &eval, 1123 const Fortran::parser::OpenMPConstruct &ompConstruct) { 1124 1125 std::visit( 1126 common::visitors{ 1127 [&](const Fortran::parser::OpenMPStandaloneConstruct 1128 &standaloneConstruct) { 1129 genOMP(converter, eval, standaloneConstruct); 1130 }, 1131 [&](const Fortran::parser::OpenMPSectionsConstruct 1132 §ionsConstruct) { 1133 genOMP(converter, eval, sectionsConstruct); 1134 }, 1135 [&](const Fortran::parser::OpenMPSectionConstruct §ionConstruct) { 1136 genOMP(converter, eval, sectionConstruct); 1137 }, 1138 [&](const Fortran::parser::OpenMPLoopConstruct &loopConstruct) { 1139 genOMP(converter, eval, loopConstruct); 1140 }, 1141 [&](const Fortran::parser::OpenMPDeclarativeAllocate 1142 &execAllocConstruct) { 1143 TODO(converter.getCurrentLocation(), "OpenMPDeclarativeAllocate"); 1144 }, 1145 [&](const Fortran::parser::OpenMPExecutableAllocate 1146 &execAllocConstruct) { 1147 TODO(converter.getCurrentLocation(), "OpenMPExecutableAllocate"); 1148 }, 1149 [&](const Fortran::parser::OpenMPBlockConstruct &blockConstruct) { 1150 genOMP(converter, eval, blockConstruct); 1151 }, 1152 [&](const Fortran::parser::OpenMPAtomicConstruct &atomicConstruct) { 1153 genOMP(converter, eval, atomicConstruct); 1154 }, 1155 [&](const Fortran::parser::OpenMPCriticalConstruct 1156 &criticalConstruct) { 1157 genOMP(converter, eval, criticalConstruct); 1158 }, 1159 }, 1160 ompConstruct.u); 1161 } 1162 1163 void Fortran::lower::genThreadprivateOp( 1164 Fortran::lower::AbstractConverter &converter, 1165 const Fortran::lower::pft::Variable &var) { 1166 fir::FirOpBuilder &firOpBuilder = converter.getFirOpBuilder(); 1167 mlir::Location currentLocation = converter.getCurrentLocation(); 1168 1169 const Fortran::semantics::Symbol &sym = var.getSymbol(); 1170 mlir::Value symThreadprivateValue; 1171 if (const Fortran::semantics::Symbol *common = 1172 Fortran::semantics::FindCommonBlockContaining(sym.GetUltimate())) { 1173 mlir::Value commonValue = converter.getSymbolAddress(*common); 1174 if (mlir::isa<mlir::omp::ThreadprivateOp>(commonValue.getDefiningOp())) { 1175 // Generate ThreadprivateOp for a common block instead of its members and 1176 // only do it once for a common block. 1177 return; 1178 } 1179 // Generate ThreadprivateOp and rebind the common block. 1180 mlir::Value commonThreadprivateValue = 1181 firOpBuilder.create<mlir::omp::ThreadprivateOp>( 1182 currentLocation, commonValue.getType(), commonValue); 1183 converter.bindSymbol(*common, commonThreadprivateValue); 1184 // Generate the threadprivate value for the common block member. 1185 symThreadprivateValue = 1186 genCommonBlockMember(converter, sym, commonThreadprivateValue); 1187 } else { 1188 mlir::Value symValue = converter.getSymbolAddress(sym); 1189 symThreadprivateValue = firOpBuilder.create<mlir::omp::ThreadprivateOp>( 1190 currentLocation, symValue.getType(), symValue); 1191 } 1192 1193 fir::ExtendedValue sexv = converter.getSymbolExtendedValue(sym); 1194 fir::ExtendedValue symThreadprivateExv = 1195 getExtendedValue(sexv, symThreadprivateValue); 1196 converter.bindSymbol(sym, symThreadprivateExv); 1197 } 1198 1199 void Fortran::lower::genOpenMPDeclarativeConstruct( 1200 Fortran::lower::AbstractConverter &converter, 1201 Fortran::lower::pft::Evaluation &eval, 1202 const Fortran::parser::OpenMPDeclarativeConstruct &ompDeclConstruct) { 1203 1204 std::visit( 1205 common::visitors{ 1206 [&](const Fortran::parser::OpenMPDeclarativeAllocate 1207 &declarativeAllocate) { 1208 TODO(converter.getCurrentLocation(), "OpenMPDeclarativeAllocate"); 1209 }, 1210 [&](const Fortran::parser::OpenMPDeclareReductionConstruct 1211 &declareReductionConstruct) { 1212 TODO(converter.getCurrentLocation(), 1213 "OpenMPDeclareReductionConstruct"); 1214 }, 1215 [&](const Fortran::parser::OpenMPDeclareSimdConstruct 1216 &declareSimdConstruct) { 1217 TODO(converter.getCurrentLocation(), "OpenMPDeclareSimdConstruct"); 1218 }, 1219 [&](const Fortran::parser::OpenMPDeclareTargetConstruct 1220 &declareTargetConstruct) { 1221 TODO(converter.getCurrentLocation(), 1222 "OpenMPDeclareTargetConstruct"); 1223 }, 1224 [&](const Fortran::parser::OpenMPThreadprivate &threadprivate) { 1225 // The directive is lowered when instantiating the variable to 1226 // support the case of threadprivate variable declared in module. 1227 }, 1228 }, 1229 ompDeclConstruct.u); 1230 } 1231